High Throughput Sequencing Revolution Reveals Conserved Fundamentals of U-indel Editing

Abstract

Among Euglenozoans, mitochondrial RNA editing occurs in the diplonemids and in the kinetoplastids that include parasitic trypanosomes. Yet U-indel editing, in which open reading frames (ORFs) on mRNAs are generated by insertion and deletion of uridylates in locations dictated by guide RNAs, appears confined to kinetoplastids. The nature of guide RNA and edited mRNA populations has been cursorily explored in a surprisingly extensive number of species over the years, although complete sets of fully edited mRNAs for most kinetoplast genomes are largely missing. Now, however, high throughput sequencing technologies have had an enormous impact on what we know and will learn about the mechanisms, benefits, and final edited products of U-indel editing. Tools including PARERS, TREAT, and T-Aligner function to organize and make sense of U-indel mRNA transcriptomes, which are comprised of mRNAs harboring uridylate indels both consistent and inconsistent with translatable products. From high throughput sequencing data come arguments that partially-edited mRNAs containing “junction regions” of non-canonical editing are editing intermediates, and conversely, arguments that they are dead-end products. These data have also revealed that the percent of a given transcript population that is fully or partially edited varies dramatically between transcripts and organisms. Outstanding questions that are being addressed include the prevalence of sequences that apparently encode alternative ORFs, diversity of editing events in ORF termini and 5′ and 3′ untranslated regions, and the differences that exist in this byzantine process between species. High throughput sequencing technologies will also undoubtedly be harnessed to probe U-indel editing’s evolutionary origins.

Graphical abstract

The functions and mechanisms of U-indel editing of mRNA in kinetoplastid mitochondria exhibit similarities and differences across species. High throughput sequencing efforts are the key to new inquiry, including exploration of the origins and evolution of this process.

Introduction

A radical and unique type of RNA editing occurs in the mitochondria of the kinetoplastid protozoa (Aphasizheva & Aphasizhev, 2016; Hashimi, Zimmer, Ammerman, Read, & Lukeš, 2013; Read, Lukeš, & Hashimi, 2016). Uridylates (U) are inserted or deleted in targeted locations in mRNAs in a process called U-indel editing, using largely trans-acting mitochondrially-encoded template RNAs called guide RNAs (gRNAs) to direct the location and nature of the modification (Figure 1A). The insertion and deletion of Us results in the repair of frameshifts, creation of stop and start codons, and often the generation of complete open reading frames compared to the “cryptogenes” in the mitochondrial genome. Thus, the RNA editing that takes place in the kinetoplastid mitochondrion is far more drastic than the modification of one nucleotide that defines most other types of mRNA editing (Licht & Jantsch, 2016; Nisbet, Kurniawan, Bowers, & Howe, 2016; Takenaka, Zehrmann, Verbitskiy, Härtel, & Brennicke, 2013; Valach, Moreira, Hoffmann, Stadler, & Burger, 2017; Yablonovitch, Deng, Jacobson, & Li, 2017).

Figure 1. Features of U-indel editing.

AU-indel editing is a process of insertion and deletion events guided by gRNAs that contain both regions that anneal to their cognate mRNAs, and regions that will direct editing of adjacent sequence. Image from Read et al. (2016). B. Schematic depicting the associations of the RNA Editing Mediator Complex (REMC) with the Guide RNA Binding Complex (GRBC), and other proteins to form the RNA Editing Substrate Binding Complex, or RESC. Once these protein players have assembled with mRNA and the necessary gRNA, interaction with the catalytic components of editing, the RNA Edited Core Complexes (RECCs) can proceed. GAP1/2 are gRNA binding proteins within GRBC, and the brown dotted-outline circles represent RESC proteins MRB7260 and MRB10130 that are not strictly REMC or GRBC components (Read et al., 2016; McAdams, Simpson, Chen, Sun, & Read, 2018). C. Depiction of different types of mRNA molecules with origins in a single pan-edited cryptogene. Extreme 5′ and 3′ regions of the transcript never get edited; the 3′ never edited region is the binding site of the first gRNA of the pan-editing process. Pre-edited sequence exists in regions yet to be edited in order to generate an mRNA encoding a canonical protein product. The bracketed forms of the mRNA represent states that are consistent with editing in progress. In these examples, the green color exists at the boundary of the editing process (junction sequence) and the molecule may represent an editing intermediate that will continue to undergo editing to eventually achieve the fully-edited state shown at the bottom. The unbracketed molecules are examples of regions of alternative editing that are not in a junction position. The alternatively edited region within the canonically edited region would result in product in which the edited region contains non-canonical codons and/or a frameshift, or else a mis-edited product. The region of non-canonical editing upstream of the main progression of editing represents a product that is likely mis-edited.

Among Euglenozoan major lineages, mitochondrial RNA U modification is absent among euglenids but observed in the diplonemids and kinetoplastids (Faktorova, Dobakova, Pena-Diaz, & Lukeš, 2016). Basic knowledge of U-appendage RNA editing (causing insertion only) between spliced fragments of mitochondrial mRNAs and rRNAs in diplonemids is only now beginning to emerge (Valach et al., 2017). In contrast, an intense focus on U-indel editing in kinetoplastids has already led to admirable progress in elucidating the mechanism whereby it occurs, and the enzymes and protein factors involved. Briefly, U-indel editing requires an extensive collection of molecular complexes, including three related catalytic RNA Editing Core Complexes (RECCs; a.k.a., 20S editosomes) that transiently interact with an incompletely defined RNA Editing Substrate Binding Complex (RESC; a.k.a. MRB) (Figure 1B). Several recent U-indel editing reviews (Aphasizheva & Aphasizhev, 2016; Hashimi et al., 2013; Read et al., 2016) focus on these protein complexes responsible for U-indel editing and rely almost exclusively on the model kinetoplastid, Trypanosoma brucei. Here, we examine the nature of U-indel edited transcriptomes (i.e., the edited sequences themselves) in a wide range of kinetoplastids, with a focus on the enormous impact and promise of high throughput sequencing technologies now being applied to this byzantine process. Application of these technologies is well underway, as demonstrated by the development of three bioinformatics tools dedicated to U-indel genome analysis, developed by independent teams (Carnes et al., 2017; Gerasimov et al., 2018; Simpson, Bruno, Bard, Buck, & Read, 2016). Indeed, these and other high throughput sequencing approaches (Aphasizheva et al., 2014; Kirby, Sun, Judah, Nowak, & Koslowsky, 2016; Koslowsky, Sun, Hindenach, Theisen, & Lucas, 2014; Madina et al., 2014) have transformed our ability to address U-indel editing’s most fascinating paradoxes, including its very existence.

EARLY STUDIES MAP THE BASIC U-INDEL EDITED TRANSCRIPT LANDSCAPE

The kinetoplastid mitochondrial transcriptome is exceedingly complex and contains RNAs transcribed from two types of mitochondrial circular DNA molecules, the maxicircles and the minicircles (Jensen & Englund, 2012; Povelones, 2014). Dozens of nearly identical maxicircles encode the mRNAs, some or all of which (depending on the organism) require U-indel editing to generate open reading frames (ORFs). Maxicircles can encode three classes of mRNAs: 1) never edited mRNAs that are correctly encoded in the genome, 2) minimally edited mRNAs requiring editing within small domains, and 3) pan-edited mRNAs requiring editing throughout their lengths and for which editing generates the entire ORF. The vast majority of gRNAs are encoded on the minicircles that cluster into hundreds of sequence classes (Kirby et al., 2016; Koslowsky et al., 2014). Minicircles also encode small RNAs that appear to play a role in gRNA 3′ end processing (Suematsu et al., 2016) and await further characterization. To this day, complete sets of fully edited mRNAs for kinetoplastid genomes are largely missing (Figure 2), and uncertainties abound even in well-studied genomes (e.g. Read, Wilson, Myler, & Stuart, 1994). Finally, very few studies have targeted gRNA or small RNA populations (Aphasizheva et al., 2014; Kirby et al., 2016; Koslowsky et al., 2014; Madina et al., 2014).

Figure 2. Current state of investigation of U-indel edited sequences from maxicircle cryptogenes among the kinetoplastids.

Included studies are those in which RNA and/or DNA sequence is provided in text, figures, or supplementary material, that describe or compare pre-, partially-, or fully-edited sequences of maxicircle cryptogenes. Unless indicated with the half-moon symbol, RNA has been sequenced directly, cDNA has been cloned and sequenced, or amplified PCR products from cDNA templates have been cloned and sequenced. The structure of the tree is approximately based on models presented in Lukeš et al. (2018). Numbers denote the reference(s) for the analysis/es: 1. Merzlyak, Zakharova, & Kolesnikov (2001); 2. Benne et al. (1986); Feagin, Shaw, Simpson, & Stuart (1988); Maslov, Avila, Lake, & Simpson (1994); Shaw, Feagin, Stuart, & Simpson (1988); van der Spek et al. (1988); Yasuhira & Simpson (1995); 3. Gerasimov et al. (2018); 4. Bessolitsyna, Fediakov, Merzliak, & Kolesnikov (2005); Kolesnikov, Merzliak, Bessolitsyna, Fediakov, & Shoenian (2003); Merzlyak et al. (2001); 5. Kolesnikov et al. (2003); 6. Blom et al. (1998); 7. Bhat, Myler, & Stuart (1991); Feagin, Shaw, et al. (1988); Maslov et al. (1992); Shaw, Campbell, & Simpson (1989; Shaw et al. (1988); Souza, Myler, & Stuart (1992); Thiemann, Maslov, & Simpson (1994) 8. Maslov (2010); 9. Kolesnikov et al. (2003); Nebohácová, Kim, Simpson, & Maslov (2009); 10. Ramírez, Puerta, & Requena (2011) 11. Kolesnikov et al. (2003); 12. Merzlyak et al. (2001); 13. Landweber, Fiks, & Gilbert (1993); Landweber & Gilbert (1993); Maslov et al. (1994) 14. Landweber & Gilbert (1993); Maslov et al. (1994); 15. Kolesnikov et al. (2003); Landweber & Gilbert (1993); Merzlyak et al. (2001); 16. Landweber & Gilbert (1993); 17. Maslov, Hollar, Haghighat, & Nawathean, (1998); Maslov, Nawathean, & Scheel (1999); Nawathean & Maslov (2000); 18. Gerasimov, Kostygov, Yan, & Kolesnikov (2012); Kolesnikov et al. (2003); 19. Gerasimov et al. (2012); 20. Gerasimov et al. (2012); 21. Aravin, Yurchenko, Merzlyak, & Kolesnikov (1998); Kolesnikov et al. (2003) 22. Maslov et al. (1994); 23. Kim, Teixeira, Kirchhoff, & Donelson (1994); Ochs, Otsu, Teixeira, Moser, & Kirchhoff (1996); 24. Ruvalcaba-Trejo & Sturm (2011); Shaw, Kalem, & Zimmer (2016); Thomas, Martinez, Westenberger, & Sturm (2007); Westenberger et al. (2006); 25. Ruvalcaba-Trejo & Sturm (2011); Westenberger et al. (2006); 26. Avila et al. (2003); Gerasimov et al. (2018); Maslov et al. (1994); Ruvalcaba-Trejo & Sturm (2011); Westenberger et al. (2006); 27. Blom et al. (1998); 28. Read, Fish, Muthiani, & Stuart (1993); Read, Jacob, Fish, Muthiani, & Stuart (1993); 29. Abraham, Feagin, & Stuart (1988); Benne et al. (1986); Bhat, Koslowsky, Feagin, Smiley, & Stuart (1990); Carnes et al. (2017); Corell, Myler, & Stuart (1994); Decker & Sollner-Webb (1990); Feagin, Abraham, & Stuart (1988); Feagin, Jasmer, & Stuart (1987); Feagin, Shaw, et al. (1988); Feagin & Stuart (1988); Kirby & Koslowsky (2017); Koslowsky, Bhat, Perrollaz, Feagin, & Stuart (1990); Read, Myler, & Stuart (1992); Read, Wilson, et al. (1994); Shaw et al. (1988); Simpson et al. (2017); Simpson et al. (2016); Souza et al. (1992); Souza, Shu, Read, Myler, & Stuart (1993); 30. Greif, Rodriguez, Reyna-Bello, Robello, & Alvarez-Valin (2015); 31. David et al. (2015); 32. Maslov & Simpson (1994); Lukeš et al. (1994); 33. Blom et al. (1998). The identities of the cultures described as Crithidia sp. KVI, Herpetomonas sp TCC263 (Kolesnikov et al., 2003), and W. inconstans (Merzlyak et al., 2001) were not known and thus were not placed on the tree.

The edited sequences of maxicircle-encoded mRNAs were initially determined by studies in the late 1980’s and early 1990’s, mainly in T. brucei and Leishmania tarentolae, which employed Sanger sequencing of multiple cDNA libraries. In some cases, findings were confirmed or extended by direct RNA sequencing (Figure 2 and references therein). To define consensus edited sequences of pan-edited mRNAs, cDNA libraries were initially interrogated by 3′ RACE, using a 5′ primer corresponding to pre-edited mRNA sequence near the predicted 5′ ends of a given gene. Alignment of dozens to nearly 100 cDNA clones per gene revealed a conserved feature of the editing process. Namely, the ubiquitous recovery of mRNAs with unedited 5′ ends and 3′ ends containing progressively edited sequence in all studies demonstrated that editing proceeds in a general 3′ to 5′ direction on the mRNA (Figure 1). The consensus edited sequence of the 3′ end was identified from these partially edited cDNA clones. These 3′ consensus edited regions within cDNAs were usually bounded at their 5′ ends by variable non-consensus edited sequences of differing lengths that were termed junctions (Koslowsky, Bhat, Read, & Stuart, 1991; Figure 1A, C). Junction regions are 5′ bounded by pre-edited sequence. Consensus edited sequence present at the 3′ ends of a majority of cDNA clones was then used to construct a 3′ primer for 5′ RACE and generation of a second cDNA library. This process was repeated iteratively until a 5′ edited consensus sequence was determined. Early studies also revealed that edited mRNAs typically contain a poly(A) tail with sporadic and non-consensus U insertions, and that pan-edited mRNAs possess short never edited regions on their extreme ends. The lengths of never edited regions range from 25-40 nt on the 5′ end and are approximately 15 nts on the 3′ end. The function of the 3′ never edited region was revealed upon the identification of gRNAs: it is the region to which the first gRNA anchors; the presence of the 5′ never edited region remains mysterious.

Conventional analyses of edited mRNAs and gRNAs also hinted at the mechanisms and regulation of editing. For example, northern blot analyses demonstrated developmentally regulated accumulation of some edited mRNAs between insect procyclic and mammalian bloodstream stage T. brucei (Corell et al., 1994; Feagin et al., 1987; Feagin & Stuart, 1988; Koslowsky et al., 1990; Read et al., 1992; Souza et al., 1992), and T. congolense (Read, Stankey, Fish, Muthiani, & Stuart, 1994). For CR4 and ND7 mRNAs, differential editing appeared to be controlled at the level of 3′ to 5′ editing progression, as full length consensus sequences could be determined only in bloodstream stages, whereas 3′ partially edited cDNAs were acquired from both stages (Corell et al., 1994; Koslowsky et al., 1990). Similarly, northern blotting of specific gRNAs revealed increased abundances in life cycle stages in which edited mRNAs were decreased, suggesting that editing leads to gRNA destruction (Koslowsky, Riley, Feagin, & Stuart, 1992; Riley, Corell, & Stuart, 1994). Finally, sequencing of cDNA libraries revealed alternatively edited mRNA populations that could lead to production of more than one protein, as discussed in greater depth below (Ochsenreiter & Hajduk, 2006; Read et al., 1992; Read, Wilson, et al., 1994). Despite these findings, practical collection limits in the numbers of cDNA and gRNA sequences confined the strength of these conclusions. In addition to the extensive T. brucei and L. tarentolae investigations, U-indel editing of mitochondrial transcripts has been analyzed to varying extents in an astounding 34 different kinetoplastid species (Figure 2). For a majority of these, the study’s target was one or a very few gene products, and sometimes editing was inferred by comparison of the DNA sequence of the species in question with the edited RNA from another species. High throughput sequencing promises to rapidly and significantly expand our understanding of the process of U-indel editing as well as the breadth of and variations in U-indel editing between wide-ranging kinetoplastids.

HARNESSING HIGH-THROUGHPUT SEQUENCING

High throughput sequencing technologies have exponentially increased sequence numbers obtainable in a single experiment, thus allowing the study of the whole mitochondrial transcriptome and the investigation of previously unanswerable questions about U-indel RNA editing. Newly-available U-indel read population characterization methodologies have already enriched our understanding of these transcriptomes and will likely be used by more investigators in the future to answer distinct questions.

The dedicated tools developed to organize and describe U-indel mRNA transcriptomes are T-Aligner (Gerasimov et al., 2018), PARERS (Pipeline Analyzing RNA Editing RNA Sequencing, (Carnes et al., 2017)), and TREAT (Trypanosome RNA Editing Alignment Tool, (Simpson et al., 2016)). The tools were designed to answer distinct questions, so each program has unique characteristics, outputs, and strengths (summarized in Table 1). Each of the three programs utilizes high throughput sequencing of cDNA, aligning or otherwise organizing sequences with highly variable U content. This is accomplished by masking the Ts of both reference and reads during mapping and retaining Ts during subsequent analysis. TREAT and T-Aligner are available as open source programs; PARERS is currently not open source. Several other groups are using high throughput sequencing to collect and analyze small RNAs of the mitochondrial transcriptome; however, as they are not edited, their content has mainly been reported directly. No programs for analysis of small RNA populations have been published to date.

Table 1.

Comparison of analytical tools that have been created for transcriptomic analysis of U-indel genomes.

Approach	PARERS	TREAT	T-Aligner
Citation	Carnes et al. (2017)	Simpson et al. (2016)	Gerasimov et al. (2018)
Open Source (Program)	No (Unknown)	https://github.com/ubccr/treat (Go)	https://github.com/jalgard/T-Aligner3 (C++)
Input	Pooled library, consisting of individual libraries of PCR products of interest of between ~ 180-300 nt, including Illumina indexing and sequencing adaptors	PCR products of target (gene) of interest, less than 550 bp including Illumina indexing and sequencing adaptors, with consistent 5′ and 3′ boundaries of editing	Library of cDNA fragments generated from poly(A)-enriched mitochondrial RNA, with insert sizes of 300-400 nt, although technically, PCR products could also be used as input
Templates Required	Pre-edited and fully-edited sequences for each amplified region of analysis	Pre-edited and fully-edited sequences for each mRNA analyzed, and optionally, one or more alternatively fully-edited sequences	DNA reference for each maxicircle-encoded mRNA or cryptogene, plus 50-100 nt flanking sequence on each reference
Sequencing Strategy	50 nt single-end (MiSeq scale)	300 nt paired-end (MiSeq scale)	250 or 300 nt paired-end (MiSeq or HiSeq scale)
Output	Pooled output is separated by gene regions of interest, and then each gene’s reads are placed into either pre-edited, canonically fully-edited, or partially edited categories. User then utilizes these read files with their own downstream analysis, with the partially-edited category yielding the most telling information for mechanistic studies.	Each read covers either the entire gene or a large portion of edited sequence. The pre-edited, junction, and canonically-edited regions of each read are identified. Population-wide measures of the junction regions (start and end sites, lengths, sequences) can be compared across samples by a web-based user interface producing constrainable histograms and tables of amplicon-specific data. Individual sequences can be viewed and compared to pre-edited and fully edited sequences. Data can be further analyzed using accompanying code written in R.	Individually, reads for each cryptogene are mapped. The location of these mapped reads on the reference, and details about their editing are presented in multiple T-Aligner graphical outputs. Sequence file output includes reconstructed ORFs in nucleotide and amino acid files, along with the level of read support for each reconstruction. Finally, editing pathways of up to three of the output ORFs can be mapped on the graphical output.

Lengths of PCR products or cDNA fragments used by TREAT or T-Aligner can be larger as sequencing technologies develop the capacity to generate longer reads.

T-Aligner was designed to determine edited ORFs in previously unexplored kinetoplastids, and for comparison of mitochondrial transcriptomes across evolutionarily divergent Kinetoplastea members (Gerasimov et al., 2018). Of the three, T-Aligner is the only program that generates full-length final coding sequences de novo, solely from high-throughput transcriptome reads. It is also the only program in which there is no PCR amplification step, eliminating a potential source of bias. Reads in all states of editing are mapped to their reference cryptogenes obtained from maxicircle genomic DNA sequencing of the investigated organism. Open reading frames are then reconstructed from overlap graphs of read alignments using a breadth-first search algorithm. The longest open reading frames starting at each possible start codon are collected. T-Aligner is the optimal tool in two instances. The first instance is when sequencing U-indel transcriptomes besides L. tarentolae and T. brucei, for which we lack the full complement of experimentally-verified edited sequences. The second is in cases in which the experiment in question requires analysis of all of the mRNAs, rather than a few reporter transcripts.

Once the sequence of a cryptogene and at least one of its fully edited products is known, both TREAT and PARERS can be used to interrogate its editing parameters. PARERS was designed to interrogate the editing site specificities of the three subtypes of T. brucei RECCs (Carnes et al., 2017). PARERS permits examination of the first regions of an mRNA that become edited (the most extreme 3′ edited region). The downstream analysis of sequences is not built into PARERS; methods to do so are user-generated. PARERS separates 50 nt single-end reads of PCR amplified pooled libraries into pre-edited, fully edited and “unmatched” groups, using a string match to regions bound by the gene-specific primers used to generate the library, and segregate sequences by gene. The unmatched group contains all partially edited sequences, including those that do and do not contain junctions, which can be further examined by the user. The use of short reads and primers to many genes makes this program ideal for examining changes early in editing across many transcripts selected by the user.

TREAT was designed to assess the 3′ to 5′ progression of editing for a given transcript (RM Simpson et al., 2016). It determines reads that align to pre-edited, fully edited, or alternatively edited references. For partially edited sequences, it defines the extent of canonical fully edited sequence and the boundaries and characteristics of any junction present. Downstream analytical scripts allow the user to compare multiple parameters between samples such as 1) locations of intrinsic editing pause sites, 2) editing pause sites exacerbated in an experimental group compared to control, 3) the lengths, beginning and end sites, and sequences of junctions, and 4) plausible editing paths through sub-regions of a given transcript. These parameters can be evaluated relative to the positions of cognate gRNAs to generate mechanistic hypotheses. (McAdams et al., 2018; Rachel M. Simpson et al., 2017; RM Simpson et al., 2016). In contrast, T-Aligner is not able to assess where each read falls in the context of the full-length transcript sequence; thus, it reveals junction variation in aggregate with other variation but cannot examine specific junction sequence in the context of a full-length sequence. In summary, TREAT is the ideal tool to use when asking questions about underlying mechanisms of editing and when interrogating new factors for their roles in facilitating the initiation or progression of RNA editing.

KNOWLEDGE GAINED FROM HIGH THROUGHPUT SEQUENCING METHODOLOGIES

Features of the steady-state U-indel transcriptome

Early conventional sequencing studies hinted at several features of the kinetoplastid mitochondrial transcriptome that have been evaluated and largely confirmed in recent high throughput sequencing studies. The number and length of junction containing sequences reported in studies utilizing conventional sequencing (Ammerman, Presnyak, Fisk, Foda, & Read, 2010; Benne et al., 1986; Bhat et al., 1990; Corell et al., 1994; Feagin, Abraham, et al., 1988; Koslowsky et al., 1990, 1991, 1992; Read et al., 1992; Read, Wilson, et al., 1994; Souza et al., 1992, 1993) suggested that the presence of U-indel modifications other than those consistent with a canonical final product is extensive. Pairing these results with observations from full-gene PCR studies where all edited species were simultaneously amplified (Ammerman et al., 2011, 2013, 2010; Kumar et al., 2016; Schnaufer et al., 2001), it appeared that only a small fraction of most pan-edited mRNAs are fully edited. U-indel edited mRNAs derived from high throughput sequencing reads from T. brucei, T. cruzi, Perkinsela, and Leptomonas pyrrhocoris have since demonstrated this to be more universally true (Carnes et al., 2017; David et al., 2015; Gerasimov et al., 2018; McAdams et al., 2018; Simpson et al., 2017; Simpson et al., 2016). Table 2 compares the percentage of sequences that were pre-edited, fully edited, partially edited containing junctions, and partially edited not containing junctions in some of these studies. However, given the variation in techniques used and types of sequences reported (such as partially-edited sequences only), the numbers presented are not directly comparable. Further studies must be initiated to establish the universality of features of U-indel editing shown in Table 2.

Table 2.

Comparison of relative numbers of sequences obtained from cryptogenes at various stages of editing completion.

Reference	Organism	Method	Gene	Total Sequences	Fully edited (percent)	Partially edited (percent)	Partially edited, no junction (percent)	Partially edited, with junctions (percent)	Pre edited (percent)
Koslowsky et al. (1991)	Trypanosoma brucei	conventional cloning and sequencing	ND7	30 #	0 (0)	30 (100)	4 (13)	26 (87)	0 (0)
Koslowsky et al. (1991)	Trypanosoma brucei	conventional cloning and sequencing	A6	23 #	0 (0)	23 (100)	2 (9)	21 (91)	0 (0)
Maslov et al. (1992)	Leishamania tarentolae	conventional cloning and sequencing	RPS12	26 #	0 (0)	25 (96)	12 (46)	14 (54)	1 (4)
Landweber et al. (1993)	Herpetomonas megaseliae	conventional cloning and sequencing	COIII	36 #	0 (0)	36 (100)	15 (42)	21 (58)	0 (0)
Landweber et al. (1993)	Lafontella mariadeanei	conventional cloning and sequencing	COIII	25 #	0 (0)	25 (100)	2 (8)	23 (92)	0 (0)
Maslov & Simpson (1994)	Trypanoplasma borreli	conventional cloning and sequencing	CYb	19 #	0 (0)	19 (100)	2 (10)	17 (90)	0 (0)
Maslov et al. (1998)	Phytomonas serpens	conventional cloning and sequencing	RPS12	4	4 (100)	0 (0)	0 (0)	0 (0)	0 (0)
Maslov et al. (1999)	Phytomonas serpens	conventional cloning and sequencing	A6	7	7 (100)	0 (0)	0 (0)	0 (0)	0 (0)
Maslov et al. (1999)	Phytomonas serpens	conventional cloning and sequencing	ND8	16	0 (0)	10 (66)	n/a	n/a	6 (34)
Maslov et al. (1999)	Phytomonas serpens	conventional cloning and sequencing	CR3/G3	4	3 (75)	1 (25)	n/a	n/a	0 (0)
Ammerman et al. (2010)	Trypanosoma brucei	conventional cloning and sequencing	RPS12	51	6 (12)	40 (78)	6 (15)	34 (85)	5 (10)
David et al. (2015)	Perkinsela	high-throughput + T-Aligner precursor	COI 5′ region	823	548 (67)	157 * (19)	n/a	n/a	118 (14)
David et al. (2015)	Perkinsela	high-throughput + T-Aligner precursor	A6	3755	883 (23)	1979 * (53)	n/a	n/a	893 (24)
David et al. (2015)	Perkinsela	high-throughput + T-Aligner precursor	RPS12	89	45 (51)	44 * (49)	n/a	n/a	0 (0)
Simpson et al. (2016)	Trypanosoma brucei	high-throughput + TREAT	RPS12	251,006	14,385 (6)	201,170 (80)	9,179 (4)	191,991 (77)	35,451 (14)
Simpson et al. (2016)	Trypanosoma brucei	high-throughput + TREAT	ND7 5′	798,405	46,488 (6)	502,591 (63)	25,258 (3)	477,333 (60)	249,326 (31)
Carnes et al. (2017)	Trypanosoma brucei	high-throughput + PARERS	ND7 5′	252,105 # ^ (WT)	0 (0)	109,613 (43)	n/a	≥50,031 (20)	142,492 (56)
Carnes et al. (2017)	Trypanosoma brucei	high-throughput + PARERS	A6	66,513 # ^ (WT)	0 (0)	1955 (03)	0 (0)	1955 (100)	64,558 (97)

Sequences obtained that reflect a translatable, edited consensus from start codon to stop codon are listed here as “fully edited”, regardless of whether there are editing differences in the 5′ UTR. Except for sequences obtained in Simpson et al. (2016), and for Phytomonas serpens, sequences presented in this table largely reflect fragments of an entire mRNA. Therefore, the data here are not meant to reflect exact percentages of molecules in various stages of the editing process. Rather, they reflect variation between organisms, and in the case of T. brucei sequencing, consistency regardless of technical approach. Only sequencing endeavors for which relative numbers of sequences in various editing states were reported were considered for this table.

^*In the Perkinsela study, reads were described as containing at least one instance of alternative editing; whether or not this was in a junction position was not determined.

^#in these studies, only partially edited sequences were examined.

^{^}only the first few editing sites were analyzed.

High throughput analyses of the mitochondrial transcriptomes of L. pyrrhocoris and T. cruzi revealed insights to the degree of editing of a transcriptome, in other words, the fraction of the total mRNA population for each specific transcript that is edited in any way, and how this varies across divergent members of the order Kinetoplastida (Gerasimov et al., 2018). Sequence analysis revealed a high variability in the degree of editing between (1) transcripts, (2) the 5′ and 3′ regions of a single transcript, and (3) organisms. In L. pyrrhocoris, the COII mRNA population is almost completely fully-edited (>95%) at its tiny three-site edited domain (Gerasimov et al., 2018). Pan-edited L. pyrrhocoris mRNAs mainly exhibit an expected profile, wherein a majority of reads are edited at sites contained in the extreme 3′ end of the edited region, with a typical drop-off of editing towards the 5′ end of the transcript. Because the 5′ end of an mRNA is the last part to be edited, the percentage of reads edited at the 5′-most sites can be taken as an approximate measure of the fully edited mRNA fraction. For L. pyrrhocoris pan-edited mRNAs, fully edited molecules range from only a few reads (ND4 and G4) to ~50 percent of the population (RPS12) (Gerasimov et al., 2018). The transcript-by-transcript variability in degree of editing may be matched by variability across organisms. When T-Aligner output from T. cruzi RPS12 and COIII are compared with that of L. pyrrhocoris, it is clear that a much higher fraction of reads from L. pyrrhocoris 5′ regions contain edited sequence. In fact, for dividing epimastigote stage T. cruzi, barely any full-length RPS12 and COIII mRNA molecules are translatable (Gerasimov et al., 2018)! This is also the case for procyclic stage T. brucei, as Simpson et al. (2016) found less than 6% of RPS12 and ND7 5′ edited domain cDNA reads to be canonically fully edited within their ORFs. However, in a different life stage or environment, T. cruzi or T. brucei editing of certain mRNA may be more robust (above; Shaw, Kalem, & Zimmer, 2016). In summary, the degree to which a transcript population achieves fully-edited status is highly variable by transcript, life stage, and organism, and can now be quantified. What such differences mean at the protein or functional level is a challenging question yet to be elucidated.

Insights into the generation and roles of junction regions and non-canonically edited mRNA

It is convenient to call editing a “noisy” process for lack of a more concise description for the surprisingly high ratio of non-canonical relative to canonical U-indels observed in the majority of pan-edited molecules. However, this terminology masks the evidence that in some or most cases, editing of a site or sites in a manner that does not match the canonical sequence may play a functional role.

Summarized in Figure 3 are three ways that non-canonical modifications may serve a functional role:

1. Their presence may mark dead-end products of the editing process. However, even in this case, they can be used as a point of regulation if the organism can adjust their abundance relative to canonically-edited products as needed.

2. They are essential intermediates in the progression of editing.

3. Their presence will result in alternative product(s) in addition to the canonical product.

Figure 3. Reasons for and potential fates of mRNA molecules that exhibit alternative editing patterns.

Arrows represent progression of editing in a single molecule that has been color-coded to indicate its editing state. In the first panel, the alternative (non-canonical) sequences are present only in dead-end edited products destined for degradation. In the second panel, the alternative sequences are part of a product with a region that is alternatively edited to generate an alternative ORF. In the third panel, the region with alternative editing is present in junctions that will continue to be edited until the canonical sequence is obtained.

Results of high throughput studies are clarifying the merits and limits of these hypotheses.

Although non-canonical U insertions and deletions have been found within, rather than at, the leading edge of edited sequences in Perkinsela, L. pyrrhocoris, and T. brucei (David et al., 2015; Gerasimov et al., 2018 supplemental material; Simpson et al., 2016), they are predominantly found at the leading 5′ edge of editing, in junctions. This is evidenced by previously described individual cloning and sequencing studies (including those of Table 2), recent high throughput sequencing studies in T. brucei, and in the multitude of 5′-truncated reconstructed ORFs with mis-editing within a short initial sequence length in L. pyrrhocoris, Perkinsela, and T. cruzi (David et al., 2015; Gerasimov et al., 2018; Simpson et al., 2016). That mRNAs containing junctions comprise the majority of partially edited mRNAs supports some functional role for these non-canonical editing modifications (Table 2; Ammerman et al., 2010; D. J. Koslowsky et al., 1991).

It is especially challenging to discern whether a junction-containing partially edited sequence is either a dead-end product or an intermediate (first and third options in Figure 3). Even with the ability to obtain many sequences, it is currently not possible to kinetically track the progression of editing of a single transcript in real time. Thus, the fate of any partially edited molecule, whether to undergo future editing remodeling events or else be degraded, is unknown. However, high throughput analysis of sequences containing given mis-edited segments relative to other sequences in the population may reveal those that are more suggestive of one or the other scenario (Figure 4). For example, a likely profile of a “dead end” product is a partially edited transcript with a junction containing a series of modifications that could not be productively edited without extensive reversal of multiple existing modifications (Figure 4, yellow arrows). Conversely, an intermediate product would be identifiable by the presence of other junctions that could be generated from it by single modifications that resolve to a conserved edited sequence (Fig. 4, yellow box).

Figure 4. Successive editing modifications successively convert junction sequence to canonically edited sequence.

A. The pre- and fully-edited versions of a short model sequence, found within an mRNA requiring editing, are shown. B. Example junction sequences found in this region such as those observed through high throughput sequencing. The dark blue and light blue, and violet > symbols represents editing intermediates shown in C. C. Noted junction sequences from B. have been ordered such that a progression from one junction to another can be generated by a single editing modification (underlined) from the sequence before it, beginning with the pre-edited and ending with the fully edited sequence. On the far left is a pathway where successive non-canonical modifications progressively build up, resulting in a junction that possibly can no longer be productively re-modified (delineated with bold u insertions). On the right, two paths of editing are shown, one with junctions resolving to the fully edited sequence (left) and the other showing a direct progression of editing (right). These pathways are based on observations in Simpson et al. (2017).

There is a lack of consensus regarding the nature of junctions as intermediates. Authors of the Perkinsela U-indel transcriptome sequencing projects noted that the vast majority of junction-containing reads (containing non-canonical U insertions or deletions) were present in extremely low copy number. Assuming that junction sequences that are part of a regular pathway of successive editing modifications and mRNA/gRNA realignments would appear with a certain frequency, they argued that their findings indicated that the mRNAs containing such junctions were likely dead-end products (David et al., 2015). Furthermore, high abundances of reads present at single or low copy number that contained alternative sequence were also found in the L. pyrrhocoris and T. cruzi transcriptomes (Gerasimov et al., 2018). The authors here also noted the previously described situation of too many junctions requiring multiple revisions to realistically envision them as intermediates. It is possible that the fraction of junctions representing editing intermediates may differ by species. To resolve this issue, more quantitative analyses of these non-PCR based data specifically to address this question is needed, such as quantifying variation within junction sequences and determining the frequency at which ORFs contain internal mis-edited sequence. Currently, such analyses are outside the scope of the T-Aligner software.

The hypothesis of junctions as intermediates has been investigated primarily in T. brucei. Authors of two other studies also observed great diversity in junction sequences; however, these modifications were not random as would be expected with mere ‘noise’ but rather occurred at a subset of editing sites suggesting a directed process (Carnes et al., 2017; Simpson et al., 2016). Analysis of these sequences showed many junctions that differed by single editing modifications as expected with successive intermediates (Ammerman et al., 2010; Koslowsky et al., 1992; McAdams et al., 2018; Simpson et al., 2017; Simpson et al., 2016). Further, depletion of certain RESC components that impede 3′ to 5′ editing progression simultaneously resulted in a loss of junction sequences (Simpson et al., 2017), suggesting that junction sequences are necessary for normal editing. The authors point to the ever-shifting mRNA/gRNA duplex as a potential source for thermodynamically unavoidable intermediates that could direct non-canonical modifications at consistent locations, generating junctions. The location of junctions at the 5′ leading edge of editing is consistent with this model. Expansion of order of modification analysis (Figure 4), especially under stress conditions, different life cycle stages and during depletion of editing factors will be needed to confirm these initial hypotheses.

Evidence that non-canonical editing generates alternative ORFs, as shown in the middle panel in Figure 3, has been extensively sought with some success. Alternatively edited mRNA molecules or reconstructed ORFs have been identified in T. brucei, T. cruzi, L. mexicana, L. pyrrhocoris, and Perkinsela (Gerasimov et al., 2018; Kirby & Koslowsky, 2017; Madina et al., 2014; Maslov, 2010; Ochsenreiter & Hajduk, 2006; Read et al., 1992; Read, Wilson, et al., 1994; Simpson et al., 2016). The more targeted studies have also identified the gRNAs that could direct these alternative editing events (Kirby & Koslowsky, 2017; Madina et al., 2014; Ochsenreiter & Hajduk, 2006). In the most complete pursuit of such products, identification of an alternatively edited mRNA, a potential causative gRNA, and antibody-detectable protein product arising from the T. brucei COIII cryptogene were identified (Ochsenreiter, Anderson, Wood, & Hajduk, 2008; Ochsenreiter & Hajduk, 2006). The abundance of this product, identified by conventional sequencing, is not known. Ideally, in addition to the types of evidence collected for T. brucei COIII, proof of alternative editing would also include proof of ribosome association of the alternative transcript, and that the encoded protein (identified by mass spectrometry or immunodetection) is depleted upon disruption of U-indel RNA editing. This idealized scenario faces technical hurdles. Approaches for ribosome profiling of organellar ribosomes have yet to be developed, and even protein products predicted from canonically edited mRNAs have been notoriously difficult to observe by traditional methods, possibly due to their extreme hydrophobicity (Acestor et al., 2011; Horváth, Berry, & Maslov, 2000; Horváth, Kingan, & Maslov, 2000; Panigrahi et al., 2009; Škodová-Sveráková, Horváth, & Maslov, 2015). As even the presence of a protein may not be evidence that it is functional (Ziková, Verner, Nenarokova, Michels, & Lukeš, 2017), investigations of potential product function, such as those performed by Ochsenreiter, Anderson, Wood, & Hajduk, 2008, would ultimately be necessary. Development of tools for genetic manipulation of the trypanosome mitochondrion would greatly aid these endeavors.

However, with the advent of high-throughput sequencing, examples of reading frames generated by alternative editing are compounding. Small alternative editing events that result in frameshifts, particularly near the 5′ end of the editing domain, would cause the formation of an alternative N-terminus or else an ORF with an entirely different sequence. These have been identified or inferred in T. brucei (Kirby & Koslowsky, 2017; Koslowsky et al., 2014; Madej, Niemann, Hüttenhofer, & Göringer, 2008; Madina et al., 2014; Ochsenreiter, Cipriano, & Hajduk, 2008; Simpson et al., 2016). The non-canonical L. pyrrhocoris and T. cruzi full length ORF reconstructions with highest support were those in which one or a few alternative editing events in a discrete region shift the reading frame such that the encoded protein contains a canonical central domain with an altered N- or C-terminus (Gerasimov et al., 2018). In fact, this is what was observed to generate the T. brucei COIII alternative (Ochsenreiter & Hajduk, 2006). Additionally, as represented by the ORF primarily colored green in the middle panel of Figure 3, T-Aligner reconstructed rare full-length ORFs (that utilized only one or very few reads over the divergently edited region), with long stretches of alternatively edited sequence that encodes an entirely different amino acid sequence. These regions were extensive enough that multiple gRNAs, up to seven, would potentially be required to direct their editing. The ORFs were hypothesized to be too rare to be functionally relevant, but revealed an inherent capacity of the genome to produce entirely different proteins (Gerasimov et al., 2018). The mounting evidence of alternative RNA products will undoubtedly spur the necessary technical advances to determine whether the fate of alternatively edited mRNAs is in fact translation. Additionally, with methods now in place to acquire complete expressed gRNAs populations (Kirby et al., 2016; Koslowsky et al., 2014), abundances of gRNAs responsible for alternatively edited transcripts can be compared to their canonically-guiding counterparts. These ratios in turn can be compared to the distribution of canonical and alternatively edited sequences in U-indel transcriptomic data.

High throughput sequencing as a tool to illuminate mechanism

The impact of high throughput sequencing to enhance mechanistic studies of editing, going beyond protein-protein interactions, is already evident. Studies utilizing this technology both illuminate the nature of kinetoplastid mitochondrial transcriptomes and reveal aspects of their generation. For example, the catalytic RECCs (Figure 1B) are present in three predominant types, each containing a distinct endonuclease (KREN1, KREN2, or KREN3), which cleaves mRNA prior to U insertion or deletion. Using high throughput sequencing and the PARERS tool, Carnes et al. (2017) investigated the in vivo specificity of KRENs by analyzing editing modifications across multiple RNAs in T. brucei cells with one, two or all three KRENs deleted. These studies showed that KREN2 and KREN3 cleave at U insertion sites, with differing but overlapping specificities, and that KREN1 exhibits specificity for U deletion sites. In addition, while distinct RECCs exhibited specific preferences in the types of sites at which their component endonucleases cleaved RNA, the number of Us inserted and deleted at a given location varied widely. These findings suggest that junction non-canonical U indels are unlikely to be a product of aberrant RECC action, but rather reflect variable modifications at locations that are targeted with some specificity.

Simpson et al. (2017) investigated the functions of three components of the REMC module of RESC (Figure 1B). While TbRGG2, MRB8180, or MRB8170/4160 were known to interact in vivo (Ammerman et al., 2012; Aphasizheva et al., 2014; Hashimi, Cicova, Novotna, Wen, & Lukeš, 2009; Kafková et al., 2012; Panigrahi et al., 2008; Weng et al., 2008), their precise relationships and functions were not understood. TREAT analysis revealed that the three proteins have distinct but overlapping functions in facilitating progression of editing along a transcript. TREAT also allowed confirmation, by a different methodology, of the discovery (Dixit et al., 2017) that MRB8170/4160 acts in editing initiation. Further, RPS12 mRNA partially edited sequences were analyzed in a stepwise manner to determine likely paths of editing modification in the region of editing directed by the first gRNA. In wild type cells, sequences containing or lacking junctions in this region were present almost equally. Depletion of either TbRGG2 or MRB8180 dramatically decreased the proportion of junction-containing mRNAs. That is, integral components of the editing machinery actually facilitate the production of non-canonically edited junctions, while at the same time causing a fatal editing defect. This study clarified the individual roles of members of the REMC subcomplex and provides the strongest evidence to date that junctions are essential for the progression of editing.

McAdams et al. (2018) used TREAT analysis to explore a previously unstudied component of RESC, MRB7260, and showed editing pausing patterns in MRB7260 knockdowns that were very distinct from those of other RESC components (Simpson et al., 2017). Additional impacts of silencing including grouped pauses just beyond gRNA anchor sequences, several pauses at gRNA ends, and apparent utilization of an incorrect gRNA, could also be determined by these methods. Parallel co-immunoprecipitation, glycerol gradient, and RNA immunoprecipitation studies supported a role for MRB7260 in gRNA trafficking and utilization. These studies demonstrate how high throughput analysis of editing defects in specific knockdowns support and extend traditional genetic approaches.

Two studies have employed high throughput sequencing for the characterization of RNA populations associated with the editing machinery. In the very first utilization of high throughput sequencing in a study of U-indel RNA editing, Madina et al. (2011) demonstrated differences in the gRNA populations isolated from native complexes containing RNA Editing Helicase 2 (REH2) compared with those of the RESC component MRB3010. However, the small amount of RNA present in immunoprecipitated complexes yielded incomplete gRNA coverage and low read counts for some gRNAs. Subsequently, Dixit et al. (2017) combined in vivo mRNA-protein crosslinking with high throughput sequencing to define the cohort of T. brucei mitochondrial RNAs bound by MRB8170 and MRB4160. These authors demonstrated the feasibility of obtaining 50,000-100,000 reads from a single affinity purification of a tagged mitochondrial protein. Their experiments highlighted differences in mRNA preferences of these two orthologous RNA binding proteins, activities which had otherwise been difficult to distinguish. However, the authors only examined reads that matched the pre-edited or fully edited sequences, leaving the large set of alternatively edited or partially edited reads uninvestigated. We expect that with the advent of approaches that utilize T-masking to align sequences and the growing library of partially edited intermediates as a reference, this “missing” data can be utilized and may yield additional insights.

Finally, sequencing of minicircle transcriptomes can inform the mechanisms of gRNA processing and contribute to models of U-indel editing. Suematsu et al., (2016) analyzed high throughput sequences to identify minicircle-derived RNAs transcribed in antisense orientation and develop a model for the processing and maturation of gRNAs. Recently published high throughput transcriptomes of gRNAs from both procyclic and bloodstream form T. brucei generated nearly complete gRNA coverage for most mRNAs (Kirby et al., 2016; Koslowsky et al., 2014), thereby permitting analysis of this U-indel mRNA transcriptome in the context of its corresponding gRNA population. Structural interactions between gRNAs and mRNAs can be predicted, and alternative and/or rare gRNAs can be observed that may be responsible for certain junction region sequences or alternatively edited products. For instance, some junction-containing mRNAs likely result from editing guided by a noncognate gRNA that has spuriously anchored at an aberrant location (as was shown for Leishmania tarentolae COIII and RPS12 in Sturm et al., 1992). High-throughput methods can now tell us how common this junction-generating mechanism likely is. Already, high throughput maxicircle transcriptome output in T. brucei has been directly compared to gRNA coverage maps to support models of proteins that affect either progression through a gRNA or gRNA exchange (Kirby & Koslowsky, 2017; McAdams et al., 2018; Simpson et al., 2017; Simpson et al., 2016). Moreover, early studies suggesting that gRNAs are consumed during editing were confirmed by Aphasizheva et al. (2014) by using high throughput sequencing to interrogate the gRNA population in cells in which editing was inhibited. Clearly, the relationship between the small RNA and maxicircle U-indel edited transcriptomes has great potential to reveal dynamics and variation in gRNA/mRNA interactions.

HIGH-THROUGHPUT SEQUENCING TO ILLUMINATE U-INDEL EDITING’S ORIGINS AND ADVANTAGES

The complexity of the U-indel editing process combined with its apparent confinement to one organelle in one group of organisms has fueled multiple ideas about why it exists in nature. There are multiple and not necessarily competing hypotheses regarding the origins of editing. Already, authors of studies utilizing high throughput sequencing have begun to weigh in on these possibilities (Gerasimov et al., 2018; Kirby & Koslowsky, 2017). One model postulates that the entire system emerged despite the lack of any adaptive advantage of the U-indel editing process. Under the constructive neutral evolution scenario, existing and then increasingly specialized machinery emerged to counter an original mutation followed by subsequent ones. Since the likelihood of mutation back to an original nucleotide is very low relative to that of a new mutation, the editing machinery became essential (Gray, 2012; Gray, Lukeš, Archibald, Keeling, & Doolittle, 2010; Lukeš, Archibald, Keeling, Doolittle, & Gray, 2011; Stoltzfus A., 1999).

Selective advantages of possessing U-indel editing have also been postulated. These selective advantages could root its origins, or at least some of its development to its present state among kinetoplastids, in natural selection. The ability to generate alternative products from a single protein-coding locus discussed above is one of these. Another is that U-indel editing could serve as a mechanism of change to more rapidly evolve mitochondrial genome output (Landweber & Gilbert, 1993), leading to greater “evolvability” for an organism. Its descendants will be more successful although the mechanism in question has no immediate benefit (Gommans, Mullen, & Maas, 2009; Speijer, 2008). Establishing an additional regulatory level with which to manipulate gene expression, or a level at which to fine-tune expression control, is also of potential benefit, especially across a varied life cycle (Landweber, 1992; Speijer, 2008). Another perhaps less intuitive benefit of editing is that the fragmentation of genetic information resulting from gRNAs encoded on separate molecules from the mRNA cryptogene may have a protective effect for the genome as a whole under periods of relaxed selection. For example, this could be beneficial in the T. brucei bloodstream stage, where the organism requires few mitochondrial-encoded proteins (Speijer, 2008). In fact, fragmentation has been proposed to be mathematically advantageous for maintaining genome integrity during periods of relaxed selection (Buhrman, van der Gulik, Severini, & Speijer, 2013). Finally, a situation may have arisen in which U-indel editing was necessary to counteract another evolutionary development. It has been proposed that G-quadruplexes inherent in the maxicircle genome are important to separate its replication from its transcription. In this case, G nucleotide regions of the pre-edited mRNA quadruplexed with the G nucleotides in the DNA acting as the switch between transcription and replication (Leeder, Hummel, & Göringer, 2016).

The origins and evolution of editing are entwined with the possibility that organisms are currently exploiting editing as a regulatory mechanism, a scenario in which it is possible to point to actual examples. As described above, that T. brucei edits certain mRNAs in one or another of its life stages has long been known (and recently systematically shown in (Gazestani, Hampton, Shaw, Salavati, & Zimmer, 2018)). Mimicking T. cruzi starvation within the insect gut in culture results in an increase in edited but not pre-edited forms of analyzed mitochondrial mRNAs (Shaw et al., 2016), an example of a regulatory response to an environmental stimulus. A broader understanding of such advantages of editing, as well as the diversity of editing parameters among kinetoplastids, such as the degree of editing within cryptogene loci (Aravin et al., 1998; Blom et al., 1998; Gao, Kapushoc, Simpson, Thiemann, & Simpson, 2001; Gerasimov et al., 2012; Landweber et al., 1993; Landweber & Gilbert, 1993; Maslov et al., 1994; Maslov, 2010; Merzlyak et al., 2001; Ramírez et al., 2011; Thiemann et al., 1994), will inform our perspective of U-indel editing’s origins and evolution.

Currently, probing of models of selective advantage and/or constructive neutral evolution of U-indel editing is limited by the disproportionate use of only a few select kinetoplastids as an investigative focus. However, the genome revolution is allowing the discovery and characterization of new kinetoplastids (e.g. David et al., 2015; Flegontov et al., 2013, 2016), of which there are hundreds or more (Maslov, Votýpka, Yurchenko, & Lukeš, 2013). Once a kinetoplastid’s maxicircle genome is sequenced, it will be increasingly simple to obtain its mitochondrial U-indel and gRNA transcriptomes, even if investigation of the editing complexes and their functions remain labor-intensive.

Specific examples of questions to be addressed are easy to identify. For instance, loss of some editing and minicircle complexity in L. tarentolae cultured for long time periods relative to a recent isolate was reported (Gao et al., 2001; Thiemann et al., 1994), but extensively-cultured L. mexicana did not exhibit similar loss (Maslov, 2010). With our new sequencing capabilities, this phenomenon can be further explored to determine whether loss of editing is typical for trypanosomatid species in culture, or rather reflects an extensive natural variation in editing and gRNA complexity (or transcription) among species and isolates. Similarly, comparisons of editing between the closely related T. brucei, T. congolense, and T. vivax could be performed at a more informative level (Greif et al., 2015; Kirby & Koslowsky, 2017; Read, Fish, et al., 1993). Use of high-throughput sequencing when comparing editing between environmental circumstances or life stages will eliminate the bias of examining only easy-to-analyze genes, and it may generate data of a depth that will illuminate the underlying mechanisms for these differences. Examination of more kinetoplastid genomes will reveal how extensively alternative editing resulting in multiple protein products is utilized, and whether this varies by species.

Finally, it may be possible to ultimately tease apart the impetus for U-indel editing’s emergence as being more rooted in selection or in constructive neutral evolution with the aid of these new technologies. Criteria have been proposed to test processes that result from constructive neutral evolution as opposed to selection. For instance, U-indel editing resulting from constructive neutral evolution might be expected to result in a smooth increase in complexity over time in descendants, and its distribution should not be too varied (Speijer, 2011). These sorts of comparisons can be facilitated with high-throughput sequencing of multiple maxicircle and minicircle genomes and transcriptomes. In summary, the sequencing revolution has positioned us to answer not only important mechanistic, but also fundamental questions about the role of U-indel editing across the wide range of species that utilize this fascinating RNA modification process.

Conclusion

The application of high throughput sequencing technologies to U-indel editing holds the promise of rapid and fundamental advances in our understanding of numerous aspects of this complex process, and recent publications from several laboratories demonstrate the power of this approach. Questions that can now be addressed and features whose functional significance can now be understood using these technologies and platforms include:

1. What percent of a given transcript population is fully or partially edited, and how does this parameter differ between transcripts?

2. How does editing progress along different mRNA populations? What are the intrinsic barriers to editing progression? How does intra-mRNA or mRNA-gRNA structure contribute to progression?

3. How prevalent are mis-edited junctions, and what are their features? What fraction of junctions exhibits sequences that could be easily re-modified to canonical edited sequence and what fraction spans long distances (perhaps greater than one gRNA length) that appear to be difficult to properly re-modify?

4. How prevalent are sequences that apparently encode alternative ORFs?

5. What is the extent of U-indel editing-generated 5′ UTR diversity? Future studies could also address whether this parameter correlates with ribosome association.

Importantly, analysis of the above parameters in different life cycle stages or environmental conditions will reveal the degree to which the editing process is regulated and the steps that are amenable to regulation. High throughput sequencing-based examination of cells depleted of specific editing machinery subunits or expressing mutant subunits will provide key insights into the mechanism of U-indel editing. Finally, comparison of the above features between mRNA populations from different kinetoplastids will illuminate the evolution, and perhaps the raison d’etre of a bizarre and fascinating process.